The Flyworks MCP server exposes a Model Context Protocol (MCP) interface to the Flyworks API for rapid, zero-shot lipsync video generation across a range of digital avatars, including realistic and cartoon styles. It supports creating lipsync videos from avatar video/audio inputs or from text using text-to-speech, and also creates digital human avatars from images or videos. Two primary operations are provided: create_lipsync_video_by_audio and create_lipsync_video_by_text. Both can use either an existing avatar or generate a new one from media inputs. The server supports both asynchronous and synchronous workflows: in async mode it returns a task_id for later polling; in sync mode it waits up to 10 minutes (600 seconds) for completion and downloads the result. Outputs can be saved to a user-defined path. The server runs on Python 3.8+ and requires httpx and mcp[cli], and is configurable via environment variables such as FLYWORKS_API_TOKEN and FLYWORKS_API_BASE_URL. Free trial access is offered with a token, subject to daily quotas and watermark/45-second limits; for full access contact Flyworks. Installation and usage guidance cover Claude Desktop and Cursor MCP clients, Smithery installation, and local server startup.